Pay it forward? Cooperative behaviour spreads through a group, but so does cheating

edyong — Mon, 08 Mar 2010 14:00:59 +0000

Ever wonder if acts of kindness or malice really do ripple outwards? If you give up a seat on a train to a stranger, do they go onto "pay it forward" to others? Likewise, if you steal someone's seat, does the bad mood you engender topple over to other people like a set of malicious dominoes? We'd all probably assume that the answers to both questions were yes, but James Fowler and Nicholas Christakis think they have found experimental evidence for the contagious nature of cooperation and cheating.

The duo analysed data from an earlier psychological experiment by Ernst Fehr and Simon Gachter, where groups of four volunteers had to decide how much money to put in a public pot. For every unit they chipped in, each member would get 0.4 back. So any donations represent a loss to the donor, but a gain to the group as a whole. The best way for the group to benefit would be for everyone to put in all their money, but each individual player could do even better by putting in nothing and feeding off their peers' generosity.

This "public goods game" went on for six rounds. At the end of each one, the players were told what their other comrades did, although everyone's identities were kept secret. The groups were shuffled between rounds so that players never played with each other more than once.

Fowler and Christakis found that the volunteers' later moves were influenced by the behaviour of their fellow players. Each act of generosity by an individual influenced the other three players to also give more money themselves, and each of them influenced the people they played with later. One act became three, which became nine. Likewise, players who experienced stingy strategies were more likely to be stingy themselves.

Even though the groups swapped every time, the contagious nature of generous or miserly actions carried on for at least three degrees of separation. You can see an example of one such cascade in the diagram below. Eleni contributes some money to the public pot and her fellow player, Lucas, benefits (one degree). In the next round, Lucas himself offers money for the good of the group, which benefits Erika (two degrees), who gives more when paired with Jay in her next game (three degrees). Meanwhile, the effects of Eleni's initial charity continue to spread throughout the players as Lucas and Erika persist in their cooperation in later rounds.

Fowler and Christakis put some numbers on these effects too, to see the effect that a single unit of money could have on the network. They found that for each unit of money chipped in by other group members in one round (say, Eleni), a player (Lucas) would put in an extra 0.19 units in the next round on average. Influenced by this cascade of charity, the next player in the chain (Erika) would put in an extra 0.05 units. And rather than rebounding to being selfish, these co-operators all continued to add in more money in later games, for the entire length of the experiment.

Many studies have shown that abstract concepts like happiness or ideas can spread from person to person and it may seem intuitively obvious that this happens. But in most cases, it's not clear if people are actually influencing their peers, it they're all influenced by the same environments (like an earthquake or an economic meltdown), or if they simply choose to hang around others who think and behave like they do. But the last two explanations clearly don't apply to Fowler and Christakis's data because their players were randomly grouped.

Fowler and Christakis suggest that people tend to mimic the actions of those they played with. They could be directly imitating the actions of other players, or they could be looking out for cues that tell them the 'right' or 'normal' way of behaving. Whether it's specific actions or social norms that are spreading, the result is the same - a ripple effect that causes groups of people to act in similar ways.

In this way, small changes could spread throughout an entire group. Fowler and Christakis claim that "social contagion... may play an important role in the evolution of cooperation" since these ripples of behaviour would encourage members of a community to behave similarly to each other, a scenario that fosters cooperation. This all seems like fairly rosy support for the "pay-it-forward" philosophy but there are reasons to be cautious about the duo's optimistic take on the data.

For a start, they repeatedly acknowledge that both cooperative acts and selfish ones can spread throughout a group. Even though cooperation takes centre stage in the title of their paper and in the way the results are framed, they've really shown that social contagion can be just as harmful as it is helpful.

Their claim that ripples of behaviour could help to foster cooperation is also a bit dubious. In Fehr and Gachter's original experiment, the players only ended up cooperating with one another if they could punish cheats at a small expense to themselves. Without punishment, they soon stopped working together. But Fowler and Christakis found signs of contagious behaviour with or without punishment, which suggests that this spread has very little influence on whether people end up working together for the common good.

Benedikt Herrmann, who has done similar psychological research with people around the world, points out that the original experiments were also done with Swiss students. They may have been anonymous but they probably knew that they were playing with other students from the same university. This is important because people are more likely to cooperate with others from the same social group.

In real life, similar games with strangers might take a very different turn. In the Fehr-Gachter experiments, Herrmann says, "It might be the case that the 'spreading of cooperation' only took place because the norm that one should cooperate with strangers was already properly established." Likewise, players from other countries, with different ethics of cooperation, might react in other ways. Herrmann, says, "In Russia, [during my own experiments], I was frequently asked by participants why they should cooperate with people who they don't know and who they regard as potentially malfeasant strangers."

All in all, Fowler and Christakis have put forward some interesting evidence for the spread of good behaviour. It's certainly nice to imagine our acts of kindness reverberating among groups of strangers who we'll never meet, but whether experiments with Swiss students really reflect this reality is unclear. And with bad deeds spreading with equal vigour, it's also unclear how these ripples affect real human networks.

Reference: Fowler, J., & Christakis, N. (2010). Cooperative behavior cascades in human social networks Proceedings of the National Academy of Sciences DOI: 10.1073/pnas.0913149107

More on cooperation:

edyong Mon, 03/08/2010 - 09:00

Categories

Brain and Behavior

I think this is not infectious behavior, but merely a change in the optimal strategy. The optimal strategy when you don't know what anyone is going to do is to defect. But as you get information, this changes. If you think it likely that you cooperation will prompt others to cooperate, or keep other cooperators cooperating, this becomes a more optimal strategy.

But perhaps this counts as infectious cooperation? But in that case it is no different to the waves of cooperation you get in spatial prisoners dilemmas with Tit-for-tat strategies.

Communication, Cooperation and Cheating: Quorum Sensing Drives the Evolution of Cooperation in Bacteria Tamas Czaran and Rolf Hoekstra, PLoS ONE 2009

Computationally, it was found that cheaters/misers/hoarders can only be present in a population at low rates for the emergence of a common good to be metabolically favorable. Sometimes I wonder whether this is also applicable to human society...

Intriguing. I'd very much like to see a cross-cultural comparison.

Newborn babies mirror a parent sticking out a tongue within a half hour of birth, but take about a year to mirror a yawn of a parent. Considering how visually contagious yawning is, and the negative reaction that a visually stuck-out tongue elicits, that is unexpected. Humans are extremely contagious critters, spreading speech & gestures loaded with common imaginary symbols along channels (like ners) occupies most of our uptime. Thank goodness for sleep, so far non-communicable downtime. (But Ed probably tweets in his sleep.)

Hmm. I see that the pay-forward factor was pretty close to 1/4 per generation -- would it be relevant that these were four-person groups?

DDeden: ;-P

I have played this game with people I didn't know and would never meet again.

The basic problem depending on the structure of communications allowed is to make the 80 x . 4 = 32 a focal point, when it is "easier" to see that a contribution is a waste.

People tend to play this game experimenting with some contributions, finding that their contributions were not sufficient and then recoiling.

In more game theoretic terms, the individuals fail to determine what the value of the coalition of all players is, and so do not move to the next step of figuring out how to build the coalition.

I agree with Tim Lucas, that the effect observed is already fairly cleanly explained by game theory. I think. I'd have to ask my friend who studied it...